Best European Alternatives to Vercel

Why people leave Vercel

Vercel is the best deployment experience in this market and almost nobody disputes that. A git push becomes a preview URL, the framework detection works, and the people who build Next.js also build the platform it runs on. Teams do not leave because the product is weak.

They leave once they read the infrastructure page. Vercel's own security documentation states that the platform "primarily uses Amazon Web Services (AWS)", and the region list names every region by its AWS identifier: fra1 is eu-central-1, cdg1 is eu-west-3, arn1 is eu-north-1. Vercel is a managed layer with a margin on top of the cloud you were probably trying to get away from, and the chain now contains two San Francisco-area companies instead of one Seattle one.

The second thing they read is the default. "The default location for Vercel functions is the U.S." — so a team in Rotterdam that never touched the region setting is running its server-side code in Washington, D.C. and its failover through Cleveland, Montréal, San Francisco and Portland before the traffic ever reaches Dublin.

• You did not leave AWS, you added a reseller This is the fact that changes the conversation. Vercel documents that its CDN and deployment platform run primarily on AWS and uses AWS Global Accelerator for its anycast routing. Every argument you had about Amazon holding your infrastructure still holds, with a second company between you and it. If the reason you are on this page is sovereignty, moving from AWS to Vercel scored zero.
• The compute default is Washington, D.C. Functions run in iad1 unless you change it, and the documented failover chain from iad1 goes cle1, yul1, sfo1, pdx1 before it reaches dub1 and lhr1. That is four American regions and one Canadian one ahead of the first European one. You can pin functions to fra1 or cdg1, and most teams never do, because nothing in the onboarding asks.
• The transfer clause is written wide on purpose Vercel's compliance page states that it "may transfer data to and in the United States and anywhere else in the world where Vercel or its service providers maintain data processing operations". It is certified under the EU-U.S. Data Privacy Framework and falls back on Standard Contractual Clauses. That is a lawful basis, not a location guarantee, and the two frameworks that preceded the DPF were both struck down.
• Isolation is an Enterprise product Secure Compute — private isolated environments and dedicated outgoing IP addresses — sits on Enterprise plans, and the VPC peering it enables is described as tunnels "within their AWS infrastructure". On Hobby and Pro you share the platform. That is normal for a managed service, and it means the answer to "can we isolate this workload" is a sales call rather than a setting.

What you have to replace, not just match

Vercel is three things bolted together, and which of the three you actually depend on decides whether this migration takes a weekend or a quarter.

The first is the build and host pipeline: git push, a build, a preview URL per branch, an atomic deploy and a one-click rollback. The second is the edge network: 126 points of presence, caching, image optimisation, the firewall. The third is compute — Vercel Functions, middleware, incremental static regeneration, streaming, and the framework features that only exist because Vercel runs the server.

Almost every team believes it needs the third and a surprising number do not. Run the check before you compare prices: if your application builds to static output and nothing runs at request time, deploybase will take it, and a handful of API routes puts Clever Cloud in front of it instead. If your product is built on middleware, on-demand revalidation and edge rendering, only Scaleway and Northflank are in the conversation, and you are porting rather than moving.

The alternatives compared

How each alternative compares to Vercel

What actually breaks when you switch

The build is the easy part and the pipeline is not. Vercel does framework detection, install, build, atomic deploy, cache invalidation and rollback as one invisible step. Everywhere else those are six steps in a CI file that somebody has to write and then own. Reserve a week for this even when the application itself ports in a day.

Then check every platform-specific call in the codebase. Image optimisation through next/image, edge middleware, on-demand revalidation, geolocation headers, the Vercel Blob and KV clients, the OIDC token integration — each of these is a line of code that compiles fine and behaves differently or not at all on another host. Grep for them before you quote a date, not after.

And plan the DNS cutover with the preview links in mind. Deployment URLs pasted into pull requests, design tickets and Slack threads stop resolving when the project is deleted, and a team that has been reviewing work through those links for two years has a lot of them. Keep the Vercel project alive and paid for a month after the switch, so the old links die slowly instead of all at once.

If Vercel runs on AWS anyway, does moving to a European host change anything?

Yes, and it is the only move on this page that does. On Vercel the bytes sit in an AWS region and two American companies hold contractual control of them: the one operating the data centre and the one operating the platform. Pinning your functions to fra1 changes neither.

On Scaleway the machine belongs to a French company in the Iliad group and sits in Paris, Amsterdam or Warsaw. On Clever Cloud it belongs to a company in Nantes. There is no American party in either contract, which is a different legal position rather than a better data centre.

If your objection was never legal — if you simply want lower latency or a lower bill — then this is the wrong reason to move and the friction section below is the part to read.

Can any of these run a Next.js application the way Vercel does?

Not identically, and pretending otherwise is how these migrations fail in week three. Next.js features that depend on the platform — middleware at the edge, on-demand revalidation, partial prerendering, the image optimiser, streaming server components — are implemented by Vercel against its own infrastructure.

What works elsewhere is the standalone Node output. Scaleway Serverless Containers and Northflank will both run that container, and Clever Cloud will run the Node process. You get the application. You do not get the edge middleware or the revalidation API behaving the way the documentation describes.

The honest version: a Next.js app that uses static generation and API routes ports in days. One built around the App Router's platform features is a rewrite of its delivery layer, and it is worth deciding that consciously rather than discovering it.

What happens to preview deployments?

Branch previews are the Vercel feature teams miss most, well ahead of anything technical, because they changed how review works: the designer opens a link instead of checking out a branch.

deploybase has them, with password-protected share links, which is why it leads this page despite being the narrowest tool on it. Northflank builds preview environments per branch including the backing services. Clever Cloud and Scaleway both require you to wire that up yourself in CI, which is an afternoon of pipeline work and then a thing your team owns forever.

Cost it before you move. A week of lost review velocity is more expensive than the hosting difference on any of these.

Is the UK a problem for Northflank?

It depends entirely on what your requirement says. Northflank is a London company, and since Brexit the United Kingdom is outside the EU and the EEA. Transfers there run on a UK adequacy decision, which is lawful and is also reviewed and renewable rather than permanent.

If your procurement rule says European, Northflank qualifies. If it says EU or EEA, it does not, and no configuration fixes that because the question is about the company rather than the servers.

This is worth settling before the technical evaluation, not after, because Northflank is the most capable platform on this page for containerised workloads and it is frustrating to reach that conclusion and then have legal veto it.

Which one to pick

Answer one question before you compare anything. If your build produces static output with at most a few form endpoints, deploybase takes it this week and leaves no American company in the chain, which is the cleanest outcome available on this page.

If the application needs a server at request time, the choice is between infrastructure and a platform. Scaleway gives you Paris, Amsterdam and Warsaw under a French company and hands back none of the developer experience. Clever Cloud keeps the push-to-deploy arrangement and asks you to accept a smaller edge story.

Northflank is the strongest platform here for a containerised stack with databases and per-branch environments, and the only one where the company sits outside the EU. If your requirement says European it is a serious candidate; if it says EU or EEA, it is not, and that conversation belongs at the start.

And if you came here to escape a hyperscaler, notice what the documentation already told you: Vercel primarily runs on AWS. Moving off Vercel is the first step of that escape, not the second.